Road Lane Line Detection using Machine Learning

Abstract

Localization of the vehicle regarding street paths assumes a basic part in an attempt to make the vehicle completely independent. Perception oriented street lane line detection gives a practical and minimal expense arrangement as the vehicle's co-ordinates are obtained from the location. Deep learning has gained wonderful advancement in the area of classification and identification of objects in an image. However, in the pursuit of automated navigation, it becomes especially challenging to identify the continuous road line and assessing path offset during heavy traffic or during a traffic jam. Another complication that has evolved of late is the correct identification of road lane exit point. Thus, the common objective of any model designed for lane line detection and/or lane line exit point notification is to determine the trajectory of the road lane with accuracy, efficiency and in real time. Conventional detection strategies need manual change of limitations, they deal with numerous issues and troubles and are still exceptionally immune to impedance brought about by deterring objects, brightening changes, and asphalt wear. Another challenge for road lane line detection are curves where the chances of accidents are very high. Instructions to successfully recognize the path line while on a curve and appropriately predict the traffic status to the drivers is a troublesome task for the offering assistance to achieve safe driving. Therefore, in this paper we propose a straight-curve model-based curve identification algorithm. This technique has shown good efficiency for most curved lane conditions. This paper has mainly focused on driver assistant framework engineering using image processing method. We have used a mounted camera on the front window of the car to map the path trajectory using the road lines and calculate where the vehicle is in relation to the path lines. Some other lane line detection techniques have also been presented in this paper such as deep learning network for path offset assessment and lane line identification in a heavy traffic situation, Hough transformation algorithm which directly recognizes the lane lines in hough spaces, lane division extraction and edge connecting method etc.